import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
from IPython.display import display

df = pd.read_csv("../../data/processed/processed.csv")
d = df.copy()
sns.set(style="whitegrid")

df.head()

unique_instancetype = df["instanceType"].unique()

len(unique_instancetype)

unique_instancetype_frame = pd.DataFrame(np.array(unique_instancetype))
unique_instancetype_frame.columns = ["unique_instancetype"]
display(unique_instancetype_frame)

df_instancetype = df.groupby("instanceType").nunique()
df_instancetype

is_numeric_col = d.dtypes.apply(lambda x: np.issubdtype(x, np.number))
df_features = pd.DataFrame(is_numeric_col)
df_features = df_features.rename(columns={0: "isNumeric"})
with pd.option_context("display.max_rows", None):
    display(df_features)

all_features = df_features.T.columns

numerical_features = is_numeric_col[is_numeric_col]
n = numerical_features.index
num_feat = pd.DataFrame(n)
num_feat.rename(columns={0: "Numerical Features"})

categorical = [x for x in all_features if x not in numerical_features]
cat_feat = pd.DataFrame(categorical)
with pd.option_context("display.max_rows", None):
    display(cat_feat.rename(columns={0: "Categorical Features"}))

with pd.option_context("display.max_columns", None):
    display(df[categorical].head())

num_feat = [x for x in all_features if x not in categorical]
with pd.option_context("display.max_columns", None):
    display(df[num_feat].head().fillna(0))

plt.rcParams.update({"figure.max_open_warning": 0})

def plot_bar_subplots(sub_df, width, height, bar_height):
    features = sub_df.columns
    height = height * len(features)
    fig, ax = plt.subplots(len(features), 1, figsize=(width, height))
    for i in range(len(features)):
        feature_freqs = sub_df[features[i]].value_counts(normalize=True) * 100
        feature_freq_df = pd.DataFrame(
            feature_freqs.sort_index(inplace=False, ascending=True)
        ).reset_index()
        g = sns.barplot(
            x=features[i],
            y="index",
            data=feature_freq_df,
            ax=ax[i],
            orient="h",
            dodge=False,
        )

        ax[i].set_ylabel(features[i], fontsize=20, va="center")
        ax[i].tick_params(axis="y", which="minor")
        ax[i].set_xlim(0, 100)
        ax[i].set_xlabel("Normalized Value Counts (%)", fontsize=20)
        ax[i].set_title(features[i], fontsize=30)

        def change_height(ax, new_value):
            for patch in ax.patches:
                patch.set_height(new_value)

        change_height(g, bar_height)
    fig.tight_layout()


def plot_bar(dataframe, n):
    features = dataframe.columns
    is_numeric_col = d.dtypes.apply(lambda x: np.issubdtype(x, np.number))
    df_features = pd.DataFrame(is_numeric_col)
    numerical_features = [x for x in features if df_features.T[x][0] is True]
    categorical = [x for x in features if x not in numerical_features]

    small_features = [
        feature
        for feature in categorical
        if len(dataframe[feature].unique()) < n
    ]
    large_features = [x for x in categorical if x not in small_features]
    dataframe[numerical_features] = (
        dataframe[numerical_features]
        .replace(["NA", "Variable"], "0")
        .astype("float")
    )

    ## Plots for features with numerical values
    plot_bar_subplots(dataframe[numerical_features], 30, 30, 0.6)

    ## Plots for features with many values
    plot_bar_subplots(dataframe[large_features], 30, 120, 0.6)

    ## Plots for features with less values
    plot_bar_subplots(dataframe[small_features], 40, 7, 0.6)

df_graph = df.copy()  # removing all unique values for simplification purposes
df_graph = df_graph.reset_index()
df_graph.drop(
    columns=["id", "instancesku", "usagetype", "index"], inplace=True
)
df_graph.dropna()
dd = pd.DataFrame(df_graph)
plot_bar(dd, 30)